Premature evaluation – The Spirit Level

Richard Wilkinson and Kate Pickett’s The Spirit Level is a vigorous polemic for social democracy, something we’re probably in need of as the neo-liberals recover from the 2008 experience.

Unlike most such, this one is based on data – specifically, a whole battery of socioeconomic indicators that turn out to be strongly correlated with income inequality. In fact, the paperback comes with a handy table of the R-squareds and p-values of all the indicators used, which range across life expectancy, imprisonment per capita, patents issued per capita and much else. Everywhere, it seems, more egalitarian societies tend to do better.

This observation is rather more impressive than quite a bit of the book – there’s too much back-of-a-fag-packet neuroscience of the sort that actual neuroscientists run a mile to avoid about mirror neurons and such, as well as a fair bit of 1970s-ish romanticisation of the supposedly ideal status of hunter-gatherer societies. Steven Pinker’s work on the history of violence hasn’t landed here; in places it’s almost nostalgically sweet.

The data, however, speaks for itself. It’s true that quite a few of the charts derive a lot of their correlation from a few outliers, but the outliers invariably point to the same results – specifically the United States, which reliably turns out to have truly awful results for many, many tests – and also very high inequality. Similarly, there are a whole string of statistics that are driven by a group of post-Soviet states that turn out to be dramatically unhappy, conflicted, violent, unhealthy, etc for their level of income; of course, these societies underwent a historic explosion of inequality.

Many of the results have been checked by carrying out the same analyses with the 51 US states, which gives rise to the same conclusion and another crop of interesting outliers. The states of the Deep South are reliably terrible. They are highly unequal, and they get the effects – but they are far off to the top right of the trendline. In a sense, their marginal productivity in terms of inequality is unusually high – for every extra point on the Gini coefficient, they manage to produce a sharply higher degree of suffering than the national average.

On the other hand, there’s the importance of being urban. The more metropolitan the state, the less it suffers from the impact of inequality – New York has the social problems of the average, despite being very unequal. And there’s the Alaskan question.

The Alaskan question? Many people on the left are keen on the idea of a citizens’ basic income, and oddly enough, there is one territory with one in this study. Alaska, famously, distributes its oil revenues equally among the citizenry, and is therefore the most equal society in the United States. However, it also succeeds in being reliably among the worst on every other measure you can think of. Clearly, the statecraft of Sarah Palin must have some impact, but it’s equally clear that it can’t be the whole explanation.

Unless there is some huge missing factor that invalidates the whole data set, we have to consider that this particular basic income experiment has failed to deliver the benefits of equality. Alaska is, of course, a very special and atypical place – but it’s not that different to, say, Norway, another sparsely populated, mountainous, northern territory bordering on Russia whose economy is heavily influenced by oil and gas, forestry, fishing, and metals and whose government decided to take a radical approach to the oil revenues, and where a lot of people own guns. And Norway is both very egalitarian and reliably in the very top of all the metrics in The Spirit Level.

Perhaps the answer is precisely that the Alaskan basic income is free money? Despite all the stuff about mirror neurons, etc, etc, it seems that the trade secret of equality is – equality. It takes a long time for Wilkinson and Pickett to get to this, but the difference between handing out oil windfalls and real egalitarianism is that only one of them is founded on a different balance of power between classes. A lasting reduction of income inequality must be founded in a lasting reduction in the inequality of political power – otherwise it may not last, and it may not even have much effect.

Another interesting point is that changes in relative economic success among nations seem to have little effect on human happiness or security. Obviously, a total crash will do it. But once a certain threshold level of per-capita GDP is passed, Wilkinson and Pickett argue, pushing into the G8 doesn’t change much. They therefore argue that economic growth is useless. However, they then note that a whole range of their metrics, like life expectancy, do seem to go up a percentage point or two a year in the rich nations anyway. Which sounds a lot like growth.

It might be more accurate to say that growth relative to other industrialised states is not particularly important within the normal range of variation, although in absolute terms it is. However, the chart in question is quite heavily driven by the US outlier – which suggests that the costs of enough inequality will essentially swallow all your economic growth.

Eventually, the upshot of TSL is that the world, and especially China, needs trade unions.

15 thoughts on “Premature evaluation – The Spirit Level

  1. Do the correlations reported by Wilkinson and Pickett reflect causative relationships? I haven’t read the book, but what evidence do they have that inequality causes any of the negative social outcomes it is correlated with? From what I’ve read about their work, it seems that they report some correlations with inequality, and then assert, with little evidence, that inequality is the cause.

    I would compare “The Spirit Level” to Richard Lynn and Tatu Vanhanen’s books about correlations between IQ and various indices of national development, where they argue that IQ is the cause. However, most people who think Wilkinson and Pickett’s correlations are causal would probably not think that Lynn and Vanhanen’s correlations are, and perhaps vice versa.

    Do Wilkinson and Pickett control for racial composition? It seems that in the US the correlations between mortality rate and income inequality vanish when you control for percent black: http://ideas.repec.org/p/pri/cheawb/263.html.

    Generally, I would argue that the socioeconomic outcomes of Americans are not that much different from those of Europeans if you control for race.

  2. This is simply a statement that inequality is high in [disfavoured group], therefore it is less so in the wider society if we exclude this group. ad absurdum, if you controlled for being poor, most of the problems would go away!

    Unfortunately, this is *precisely* what most of Wilkinson & Pickett’s critics do. if you exclude countries like the US and Sweden, the correlations are much weaker. to put it another way, the correlation is driven by the existence of countries (and US states) that are either relatively equal and rather nice, or relatively unpleasant and unequal.

    In this case, “control for” could be substituted with “deny existence of” without much loss of meaning.

  3. I’m pretty sure you didn’t read the criticism. First, the ‘what data set’ question is about not using the GINI. Second, speaking of exclusions that mess with the data, Hong Kong and Singapore both have very unequal societies but Japan-like stats. Which largely kills off the correlations when even just Hong Kong is put in the data. See <a href="http://spiritleveldelusion.blogspot.com/&quot; for example.

    And even fairly sympathetic reviewers don’t think the data is very robust. See here.

    Basically most of the data set comes down to Japan. Which suggests the question: is it something about 1st world Asia? And the answer appears to be ‘yes’ as Hong Kong and Singapore look like Japan in terms of results, but the US in terms of inequality. And that is the kind of data that would need answering before we could accept the Spirit Levels complete hand waving of a scientific causal factor.

  4. If there is no relationship between inequality and health outcomes or whatever in most countries, and the correlations are only due to a few outliers in the data, it makes no sense to propose a general theory that inequality causes poor health and other bad outcomes.

    As Sebastian says, Wilkinson and Pickett’s correlations are highly dependent both on which inequality index is used and on what countries are included in the comparison.

  5. Wilkinson and Pickett’s correlations are highly dependent both on which inequality index is used and on what countries are included in the comparison.

    This strikes me as being obvious, not to say trivial. If I was going to make statements about GDP, say, whether I used nominal, real, market rate, PPP rates, etc would have a significant impact on my results.

    Second, speaking of exclusions that mess with the data, Hong Kong and Singapore both have very unequal societies but Japan-like stats…no relationship between inequality and health outcomes or whatever in most countries, and the correlations are only due to a few outliers in the data

    May I introduce you to the concept of the social wage?

    You’ve basically assumed-out the possibility that the US’s healthcare nonsystem is a significant contributor to the ill effects of inequality in the United States.

  6. “This strikes me as being obvious, not to say trivial. If I was going to make statements about GDP, say, whether I used nominal, real, market rate, PPP rates, etc would have a significant impact on my results.”

    No, you aren’t speaking to magnitude, which is unlike you. If you use the standard, regularly utilized measure of inequality (GINI), the effect vanishes for most measures and nearly vanishes for almost all the others. That is *not* what you would expect from a robust, cause based correlation. You might expect it to be attenuated somewhat, but the fact that the normal measure of inequality doesn’t detect the effect suggests that the effect (to be VERY generous) either doesn’t exist or is very small. You can’t hide the nearly complete evaporation of the alleged effect under “significant impact on my results”.

    “May I introduce you to the concept of the social wage?” No you may not, as I’m well aware of it.

    You’re misusing it in any case. The Spirit Level makes dramatic claims about INEQUALITY and the effects of INEQUALITY on a huge number of different other lifestyle outcomes. The social outcomes of Japan drive nearly all the numbers (without it you drop below statistical significance in many cases, and well out of practical significance in the others).

    You reference the US health care as a significant contributor to the ill effects of inequality in the United States. This is of course begging the question, but I will presume that you mean it is a significant contributor to the *outcomes* measured in the Spirit Level. If you believe this is an unfair changing of your views, I will retract of course. This is no answer to the objection I raise when I mention Hong Kong and Singapore.

    Hong Kong and Singapore both show INEQUALITY on par with the United States. But contrary to what the Spirit Level would predict, they show Japan-like outcomes in all the other areas. So if you want to appeal to the explanatory value of INEQUALITY, there is a problem. If you want to appeal to the explanatory value of healthcare in the US compared to Hong Kong, that is fine, but then you have gotten into the area of local differences, which Wilkinson and Kate Pickett deny have the explanatory force of INEQUALITY. The book claims that even when you control for all those other things, inequality is a huge factor. But you can’t really think that if you add in Hong Kong and Singapore because A) the effect vanishes as a statistical matter, and B) those countries show exactly the wrong outcomes as predicted by the Spirit Level even if some small level of theoretical statistical significance survives, suggesting that in practical value the inequality issue, even at US levels, is swamped by other things.

    And if the health care explanation is true, there are a host of policy applications which are radically different from those suggested by the inequality explanation being true. So it would be good to figure out which one is correct.

  7. So if you want to appeal to the explanatory value of INEQUALITY, there is a problem.

    Not really. The US’s unusually social wage is a contributor to INEQUALITY. This explains yr anomaly.

  8. It explains it in Hong Kong? Singapore? How does the US’s unusual social wage contribute to inequality in Hong Kong and Singapore?

    It is as if you are skimming instead of reading, so I’ll try to be really brief. Hong Kong has the inequality of the US but the outcomes of Japan. This suggests that the inequality itself is not actually driving the outcomes in the way that the Spirit Level suggests. This also suggests that the equality of Japan is not actually driving the outcomes in the way that the Spirit Level suggests.

  9. I think the Alaska point is important. The factors which result in people receiving higher wages likely affect many of the outcomes other than merely through the income mechanism.

    John Rawls may have had something like that idea, since he sought reduce inequality in pre-tax income rather than simply redistribute with tax & transfers:
    http://www.themoneyillusion.com/?p=1850#comment-5182

  10. I bet they didn’t go to inequality that multicultural society brings by itself…
    Anyway equality is of no valor whatsoever unless people are actually equal, and that means choosing the same thing.
    I don’t understand the apparent surprise about “happiness” since it is something impossible to factor reliably. The only way it could be is by suicide rates. And at distance by mental health, birth rates, marriage rates.

  11. It explains it in Hong Kong? Singapore? How does the US’s unusual social wage contribute to inequality in Hong Kong and Singapore?

    Jesus wept. HK and Singapore’s higher social wages compensate for the effects of (income) inequality. The US is unusually low, and this amplifies them.

    Also, you might profit by actually reading my review – one of the interesting anomalies in the data is that being urban seems to do you good. More urbanised US states have better outcomes than would be predicted by their levels of income inequality. Those two are city-states. Nowhere on earth is more urban than Hong Kong.

  12. Pingback: Here’s something about Turkey, and everyone else | A Fistful Of Euros

  13. Pingback: Here’s something about Turkey, and everyone else | The Yorkshire Ranter

Comments are closed.